i read that good convergence method for BGP is to rely on IGP Convergence by declaring next-hop dead and not decreasing BGP timers because the latter will cause unstable network. i really fail to see what is the diffidence here

1 Answer 1


I am not sure I understand your question completely, you should at least mention where and what you read about this issue.

Here is a good blog entry to help learn about BGP Convergence.


The main takeaway is that the choice and configuration of the IGP (usually OSPF or ISIS) makes a large difference in the speed of eBGP convergence because BGP convergence is timer/scanner driven while the IGP such as OSPF is event driven. Events that cause very fast changes in IGP can be used to push fast convergence changes in BGP through a few means mentioned in that link.

Some people recommend changing the peering/session timers for the eBGP peers but you can't make really large improvements that way without risking large increases in CPU load, or simply hitting the minimum timers. Event driven updates are much faster in any case. Using Fast External Fallover triggered by those events (like loss of signal on an interface tied to the peering) or using BFD to improve discovery of link issues between peers will help speed up convergence in the case of a peer state change much more effectively than simply changing timers can accomplish.

Another option is Prefix Independent Convergence (PIC) which allows BGP to have a 'backup' path (assuming you have a second path for anything) for some/all network prefixes installed in the BGP RIB and FIB so that if the installed route in the RIB/FIB fails because of an unreachable next-hop address, the router can immediately use the second path while convergence takes place normally.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.